Sometimes the ideal of obtaining a representative sample is in practice obstructed by unavoidable circumstances accompanying the research. This happens when the share represented by some groups in a sample (for example, people with basic education) is smaller than the share of the total represent that they represent in reality. Most often this is because some people are unable to take part in the survey (for example, they are often not at home and interviews cannot get hold of them) or they are simply not willing to be interviewed and when the interviewer asks them they refuse. And these are not randomly occurring factors, as some groups are more likely to be missing from a research sample (for example, people who work long hours, people with higher income, or people with a lower level of education).
However, if the size of the error is not too great, it is possible to rectify these problems using weights, which means adjusting the importance of individual groups of respondents in the sample. For example, if we find that instead of making up 10% of our sample people with basic education account for only 5%, the weight of respondents with basic education in the sample can be doubled. The effect of this is similar to what would happen if their share were increased to the desired 10%. There are, of course, strict professional rules that govern the situations in which these adjustments can be made and the degree to which it is methodologically safe to weight data. It is definitely not possible, for example, to turn one respondent into ten.
In situations where circumstances make it impossible to obtain a perfectly representative sample of respondents directly in the course of interviewing, using weights is a very simple and methodologically acceptable way of correcting for a slight deviation. It brings us closer to ideal representativeness and we can thereby improve the quality of the research as a whole. For example, in the case of probability (random) sampling, weighting data is basically a necessity.